基于块稀疏的电阻抗成像算法  

Block-sparse Reconstruction for Electrical Impedance Tomography

作　　者：王琦[1,2] 张鹏程 汪剑鸣 李秀艳[1,2] 连志杰 陈庆良[3] 陈彤云 陈晓静 贺静[1,2] 段晓杰[1,2] 王化祥 

机构地区：[1]天津工业大学电子与信息工程学院,天津300387 [2]天津市光电检测技术与系统重点实验室,天津300387 [3]天津市胸科医院,天津300000 [4]天津大学电气与自动化工程学院,天津300072

出　　处：《电子与信息学报》2018年第3期676-682,共7页Journal of Electronics & Information Technology

基　　金：国家科技支撑计划重点项目(2013BAF06B00);国家自然科学基金(61601324;61373104;61402330;61405143);天津市应用基础与前沿技术研究计划(15JCQNJC01500)~~

摘　　要：该文提出一种基于自适应块稀疏字典学习的电阻抗图像重建算法,构建了分块稀疏字典,较好地保留了重建图像的细节信息;同时,将字典学习与图像重建交替进行,并将迭代重建的中间结果作为稀疏字典的训练样本,有效提高了字典学习效果。数值仿真与实验重建结果表明,新方法对电阻抗成像系统测量噪声具有较好的鲁棒性,能准确重构电导率分布图像,特别是对突变细节的准确恢复。An electrical impedance image reconstruction algorithm based on adaptive block-sparse dictionary is proposed. A block-sparse dictionary is constructed creatively, which preferably preserves the details of reconstructed images. Meanwhile, the sparsifying dictionary optimization and image reconstruction are performed alternately, and the intermediate result of the iterative reconstruction is used as the training sample of the sparse dictionary, which can effectively improve the learning effect of the dictionary. The numerical simulation and experiment results show that the patch-based sparsity method for measure noise has excellent robustness and can accurately reconstruct the conductivity distribution image, especially the precise details of mutation.

关 键 词：电阻抗层析成像 图像重建 稀疏表示 字典学习 

分 类 号：R444[医药卫生—诊断学]